Abstract: To extract dispatching rules from the ever-increasing workshop production data for guiding production scheduling tasks,a scheduling algorithm based on data mining was proposed.The minimize maximum completion time was set as the performance indicator,and a suitable scheduling sample set from the offline production data of the job shop was established.The established scheduling sample set was divided into training set and test set according to an appropriate ratio;then,the Classification and Regression Tree(CART)in the data mining algorithm was used to obtain effective scheduling knowledge from the training set,and a CART tree dispatching rule library was formed.To verify the effectiveness of the obtained dispatching rules,the obtained dispatching rules were combined with the genetic algorithm,and a genetic algorithm based on data mining and dispatching rules was designed as a scheduling algorithm to solve the job shop scheduling problem.Through the simulation and testing of different job shop classic examples,the effectiveness and superiority of the extracted dispatching rules and the scheduling algorithm were verified.

Key words: data mining, job shop scheduling, classification and regression tree, dispatching rules

摘要： 为了从与日俱增的车间生产数据中提取调度规则来指导生产调度任务,提出一种基于数据挖掘的调度算法。将最小化最大完工时间设置为性能指标,从作业车间的离线生产数据中建立合适的调度样本集;将建立的调度样本集按合适的比例分为训练集和测试集;用数据挖掘算法中的分类回归树(CART)从训练集中获取有效的调度知识,形成CART树状调度规则库;为了验证所得调度规则的有效性,将调度规则与遗传算法结合,设计了一种基于数据挖掘和调度规则的遗传算法作为调度算法来求解作业车间调度问题。通过对不同作业车间经典算例进行仿真与测试,验证了所提调度规则和调度算法的有效性与优越性。

关键词: 数据挖掘, 作业车间调度, 分类回归树, 调度规则